AI ML

Gen AI Developer/AI ML Engineer

We are looking for a GenAI/AI-ML engineer to build and own core LLM/RAG services, agentic workflows, and the integration layer that connects AI capabilities to our platform. You will work Python-first, ship streaming APIs, manage prompt lifecycles, and ensure safety, observability, and performance at scale.

Role Requirements What Were Looking For

Everything you need to know about this role responsibilities and the skills we value.

Strong Python skills FastAPI, asyncio, Pydantic; clean, tested, production-ready code.

Hands-on experience building LLM-powered applications (OpenAI, Gemini, Llama, or similar).

Experience with RAG architectures and vector databases (pgvector, FAISS, Pinecone, or similar).

Familiarity with agentic frameworks (LangChain, LlamaIndex, CrewAI, or custom implementations).

Knowledge of streaming APIs: SSE and WebSockets.

Understanding of prompt engineering, prompt versioning, and evaluation techniques.

Experience integrating REST/gRPC APIs and enterprise connectors.

Awareness of AI safety, guardrails, PII handling, and responsible AI practices.

Familiarity with observability tooling (OpenTelemetry, Grafana, Prometheus, or similar).

Good understanding of CI/CD pipelines and DevOps practices in an AI/ML context.

Build LLM/RAG services in Python (FastAPI/asyncio, Pydantic) with clean APIs and tests.

Implement agentic AI workflows tool-using agents with planning, memory, multi-step execution, and recovery/fallback paths.

Stream responses server-side from model to UI (SSE/WebSockets with retries, backoff, partial responses, and cancellation).

Build RAG pipelines: ingestion, chunking, embeddings, indexing, reranking, and grounded answers with citations (pgvector/FAISS/Pinecone).

Manage prompt versioning, templates/params, safe fallbacks, and rollbacks.

Run evaluations: hallucination/groundedness checks, regression suites for prompts and retrievers.

Implement guardrails: PII detection/redaction, content safety, and domain constraints.

Optimise for performance and cost: context reduction, caching/batching, request pacing, and rate-limit handling.

Design and ship REST/gRPC endpoints that orchestrate tools, retrieval, and post-processing (citations/formatting).

Implement and consume MCP (Model Context Protocol) tool adapters files, web, DB connectors with capability negotiation, auth/permissions, and resource limits.

Integrate vector stores (pgvector/FAISS/Pinecone) plus DB/file stores and enterprise connectors (SharePoint, Confluence, Slack, Jira).

Own AuthN/Z and tenancy: JWT/OAuth, role/tenant isolation, secrets management, and audit logging for AI actions.

Set up observability: logs, metrics, traces, and dashboards/alerts for latency, error rate, and token spend.

Manage queues/workflows for background ingestion/summarisation with idempotency and retries.

Enforce data governance: input/output validation, schema contracts, PII handling, and retention policies.

Maintain CI/CD for prompts, retrieval configs, and API changes; support blue/green or canary releases.